Prediction Market MCP Server

Give Claude, ChatGPT or Cursor the whole prediction market tape

Ask your AI what winning traders are buying, then test it

Connect Rivo to Claude, ChatGPT or Cursor and your assistant reads the same tape you see in the terminal: every whale trade on Kalshi and Polymarket, what winning traders hold right now, and the full record behind every name.

Then it acts on it. Ask it to find a trader worth copying and it ranks, shortlists and backtests them on settled markets. Ask it to try an idea and it starts a paper bot at live prices, so you see how the idea trades before any money is involved.

Fifteen tools, four guided prompts and one key, included with the subscription.

Winning traders your assistant can open right now

The month's most profitable named traders with an A or B copy grade on Kalshi and Polymarket. Ask your assistant to look into any of them and it reads the full record and backtests copying them.

Trader	Venue	Copy grade	Won-lost this month	Profit this month
thebigdon	Kalshi	A	453-32	+$256.8K
Wiseandstrong	Kalshi	A	412-71	+$247.8K
lost.buck	Kalshi	A	1,205-243	+$124.4K
Domster	Kalshi	A	98-21	+$109.5K
eventalpha26	Polymarket	A	163-0	+$3.0M
juice-fruit	Polymarket	A	489-225	+$1.8M
Abkha	Polymarket	A	70-0	+$1.6M
ArielFilipino	Polymarket	B	472-289	+$1.3M

Rebuilt every hour from Rivo's record of named traders. Every trader, by market: leaderboard.

Connect: add Rivo to your assistant in a minute

The server is remote, so there is nothing to install or run. Create a key in Settings, paste the endpoint into your client with the key as a bearer header, and the tools appear.

What you get

One endpoint. Claude Code, Claude Desktop, claude.ai connectors, Cursor, ChatGPT connectors and any agent built on the MCP SDKs.
A read key. Whale trades, positions, moves, markets, traders and backtests.
Write when you want it. Add write to place paper orders and run paper bots.

Endpoint

https://api.rivo.markets/mcp/v2

Claude Code

claude mcp add --transport http rivo-markets https://api.rivo.markets/mcp/v2 \
  --header "Authorization: Bearer rivo_live_..."

Cursor, Claude Desktop and JSON clients

{
  "mcpServers": {
    "rivo": {
      "url": "https://api.rivo.markets/mcp/v2",
      "headers": { "Authorization": "Bearer rivo_live_..." }
    }
  }
}

Get an API key Setup for each client

Read the tape: see what winning traders are doing right now

Ask what the money is doing and your assistant reads it straight from Rivo's tape across Kalshi, Polymarket, Polymarket US and Gemini, with each trader's record on every row.

What you get

get_positions. What traders with a winning record hold right now.
get_trades. The latest whale buys, each with the trader's record.
get_moves. Markets that just moved or flipped.
search_markets and get_market. Any market by name, its prices and who holds each side.

Get access

Judge traders and ideas: backtest before you bet

Have your assistant rank traders in any kind of market, open the best records and score copying them on settled markets. Holdout checks the result on a later stretch the test never saw, so a lucky month does not pass for an edge.

What you get

search_traders. Ranks traders by win rate, return or profit in any kind of market.
get_trader. One trader's full record.
backtest. Copying a trader or following a signal, scored on settled markets, with holdout.

The same backtest over REST

curl https://api.rivo.markets/v2/backtest \
  -H "Authorization: Bearer rivo_live_..." \
  -H "Content-Type: application/json" \
  -d '{"signal":{"venue":["kalshi"],"priceMin":0.05,"priceMax":0.25,"minUsd":10000},"holdout":true}'

Get access How backtests work

Run it on paper: turn an idea into a bot without spending a dollar

When an idea holds up, your assistant can trade it on paper at live prices. A paper bot copies a trader, follows a signal, or trades one market by a price rule, and its record builds from the next trade on. No tool on the server moves money.

What you set

place_paper_order. Buy or sell one side at the live price, including the venue's fee.
create_bot. Who or what the bot follows and your stake per trade.
update_bot and delete_bot. Change or stop it whenever you like.
get_paper_account. Your paper positions, by hand and by bots, with profit.

Get access About bots

Prompts that know the order

Four prompts tell the model which tools to call and in what order.

what_are_whales_buying reads what winning traders hold and the biggest recent buys.
look_into_this_market covers one market: prices, who holds each side and how they have done there.
find_a_trader_worth_copying goes from a ranking to a shortlist to a backtest with holdout, and ends on one recommendation.
test_this_idea_on_paper turns an idea into a signal, checks it on history and starts it as a paper bot.

Things to ask it

See what winning traders hold
"What are winning traders holding in the NBA right now? Group it by game, show which side the money is on, and tell me what each side costs now."
Shortlist Kalshi traders worth copying
"Rank Kalshi traders by win rate with at least 20 settled bets, open the top five, and backtest copying the best two over 90 days with holdout."
Test a longshot idea
"Backtest Polymarket buys between 5 and 20 cents of at least $10,000 with holdout. If it holds, sweep the top price from 10 to 30 cents and show where the return peaks."
Explain what just moved
"Show me the biggest market moves of the last hour and, for the top three, who was buying and at what price."
Look into one market
"Look into the Fed's next rate decision: the prices, who holds each side, and how those traders have done on Fed markets."
Run an idea on paper
"Start a paper bot that copies $50 a trade of whale buys on tennis favorites under 75 cents, and tell me how it is doing at the end of the week."

Every tool and the scope it needs

A key only ever sees the tools its scopes allow. A read key gets every read tool; write adds paper orders and paper bots. No tool moves money.
Tool	Group	What it does	Scope
get_trades	Whale money	Large trades, newest first, reaching back 30 days, with the market, side, price, dollars and the named trader's record. Each row older than an hour counts toward the 5,000 trades a day.	read
get_positions	Whale money	What traders with a winning record hold now, biggest first. Each holder counts as one row.	read
get_moves	Whale money	Sharp moves, favorite flips and whales piling in over the last day. Pass since to read only what is new.	read
search_markets	Markets	Markets by name or activity, open, settled or both, with refs for the other tools.	read
get_market	Markets	One market, with optional book, whales, call, history and trades. Each include costs one more request.	read
search_traders	Traders	Traders ranked by profit, win rate, return, volume or activity: the top 100 for any ranking. Each call counts toward the 10 trader listings a day.	read
get_trader	Traders	One trader's record, with optional positions, curve, habits and similar traders. Each different trader counts toward the 100 traders a day.	read
backtest	Testing	Score copying a signal or a trader on settled markets, with optional holdout, sweep and breakdown. Places nothing.	read
place_paper_order	Paper	Buy or sell one side on paper at the live price. Pass a requestId so a retry never places it twice.	write
get_paper_account	Paper	Paper positions placed by hand and by bots, with profit.	read
list_bots	Bots	Every bot on the account with its record.	read
get_bot	Bots	One bot's settings and record, with optional trades and activity.	read
create_bot	Bots	Start a paper bot that copies a trader, a signal, or trades one market by a price rule.	write
update_bot	Bots	Change a paper bot's settings, or pause and resume a trader or market bot.	write
delete_bot	Bots	Stop and delete a paper bot. Its record is kept.	write

Built for an assistant

Paper only. Paper orders and paper bots trade at live prices and no money moves.
Only the tools a key can use. A read key never sees the paper tools, so a model never plans around a call it cannot make.
Safe retries. A paper order carries a requestId, so a retry after a timeout returns the first order instead of placing a second.
Errors that say what to do next. Every failure carries a stable code and a hint.
Answers sized for a model. Lists return 20 rows unless asked for more, and every tool says which one to call next.

Limits

Limits are per account and shared with the REST API: 120 requests a minute, 5,000 trades older than an hour and 100 different traders a day. Each extra section of a market or trader and each extra backtest scan counts as one more request. See the rate limits and errors pages, and the developer docs for every route with samples in curl, Python and TypeScript.

Frequently asked questions

What is a prediction market MCP server?

An MCP (Model Context Protocol) server gives an AI assistant tools it can call. Rivo's server gives Claude, ChatGPT, Cursor or any MCP client the whale trades, trader records, market data and backtester behind Rivo, so you can ask it what the money is doing on Kalshi and Polymarket and have it check an idea against settled markets.

Is it local or remote?

Remote, over HTTP. There is nothing to install or run. Add the endpoint to your client with your API key as a bearer header and the tools appear.

Which clients work?

Anything that speaks MCP over HTTP with custom headers: Claude Code, Claude Desktop and claude.ai connectors, Cursor, ChatGPT connectors, and agents built on the MCP SDKs. The docs carry the exact setup for each.

Can an AI spend my money through this?

No. The server trades on paper only: paper orders and paper bots run at live prices and no money moves. A read key cannot even see the paper tools; a key with write access can place paper orders and start paper bots.

What does it cost?

It is included in the Rivo subscription, $9 a week or $15 a month. Just one trade can pay for it. The limits are per account and shared with the REST API: 120 requests a minute, 5,000 trades older than an hour and 100 traders a day.

How is a backtest's return worked out?

The way prediction market contracts settle: a winning contract pays $1 and a losing one pays nothing. A $100 stake at 20 cents makes $400 profit on a win and loses $100 on a loss. Stakes default to a flat $100 a trade so two ideas compare fairly. The methodology page has every formula.

Is there a plain REST API too?

Yes. Every tool returns the same data as a v2 API route, with the same keys and limits. Use the API when you want to script it yourself and MCP when you want an assistant to drive.

Give your assistant the tape

Whale trades, trader records, backtests and paper bots in Claude, ChatGPT or Cursor. Included with Rivo, $9 a week or $15 a month. Just one trade can pay for it.

Get access

rivomarkets

Launch terminal

Prediction Market MCP Server

Give Claude, ChatGPT or Cursor the whole prediction market tape

Get access Setup and tool reference

Ask your AI what winning traders are buying, then test it

Fifteen tools, four guided prompts and one key, included with the subscription.

Winning traders your assistant can open right now

The month's most profitable named traders with an A or B copy grade on Kalshi and Polymarket. Ask your assistant to look into any of them and it reads the full record and backtests copying them.

Trader	Venue	Copy grade	Won-lost this month	Profit this month
thebigdon	Kalshi	A	453-32	+$256.8K
Wiseandstrong	Kalshi	A	412-71	+$247.8K
lost.buck	Kalshi	A	1,205-243	+$124.4K
Domster	Kalshi	A	98-21	+$109.5K
eventalpha26	Polymarket	A	163-0	+$3.0M
juice-fruit	Polymarket	A	489-225	+$1.8M
Abkha	Polymarket	A	70-0	+$1.6M
ArielFilipino	Polymarket	B	472-289	+$1.3M

Rebuilt every hour from Rivo's record of named traders. Every trader, by market: leaderboard.

Connect: add Rivo to your assistant in a minute

The server is remote, so there is nothing to install or run. Create a key in Settings, paste the endpoint into your client with the key as a bearer header, and the tools appear.

What you get

One endpoint. Claude Code, Claude Desktop, claude.ai connectors, Cursor, ChatGPT connectors and any agent built on the MCP SDKs.
A read key. Whale trades, positions, moves, markets, traders and backtests.
Write when you want it. Add write to place paper orders and run paper bots.

Endpoint

https://api.rivo.markets/mcp/v2

Claude Code

claude mcp add --transport http rivo-markets https://api.rivo.markets/mcp/v2 \
  --header "Authorization: Bearer rivo_live_..."

Cursor, Claude Desktop and JSON clients

{
  "mcpServers": {
    "rivo": {
      "url": "https://api.rivo.markets/mcp/v2",
      "headers": { "Authorization": "Bearer rivo_live_..." }
    }
  }
}

Get an API key Setup for each client

Read the tape: see what winning traders are doing right now

Ask what the money is doing and your assistant reads it straight from Rivo's tape across Kalshi, Polymarket, Polymarket US and Gemini, with each trader's record on every row.

What you get

get_positions. What traders with a winning record hold right now.
get_trades. The latest whale buys, each with the trader's record.
get_moves. Markets that just moved or flipped.
search_markets and get_market. Any market by name, its prices and who holds each side.

Get access

Judge traders and ideas: backtest before you bet

What you get

search_traders. Ranks traders by win rate, return or profit in any kind of market.
get_trader. One trader's full record.
backtest. Copying a trader or following a signal, scored on settled markets, with holdout.

The same backtest over REST

curl https://api.rivo.markets/v2/backtest \
  -H "Authorization: Bearer rivo_live_..." \
  -H "Content-Type: application/json" \
  -d '{"signal":{"venue":["kalshi"],"priceMin":0.05,"priceMax":0.25,"minUsd":10000},"holdout":true}'

Get access How backtests work

Run it on paper: turn an idea into a bot without spending a dollar

What you set

place_paper_order. Buy or sell one side at the live price, including the venue's fee.
create_bot. Who or what the bot follows and your stake per trade.
update_bot and delete_bot. Change or stop it whenever you like.
get_paper_account. Your paper positions, by hand and by bots, with profit.

Get access About bots

Prompts that know the order

Four prompts tell the model which tools to call and in what order.

what_are_whales_buying reads what winning traders hold and the biggest recent buys.
look_into_this_market covers one market: prices, who holds each side and how they have done there.
find_a_trader_worth_copying goes from a ranking to a shortlist to a backtest with holdout, and ends on one recommendation.
test_this_idea_on_paper turns an idea into a signal, checks it on history and starts it as a paper bot.

Things to ask it

See what winning traders hold
"What are winning traders holding in the NBA right now? Group it by game, show which side the money is on, and tell me what each side costs now."
Shortlist Kalshi traders worth copying
"Rank Kalshi traders by win rate with at least 20 settled bets, open the top five, and backtest copying the best two over 90 days with holdout."
Test a longshot idea
"Backtest Polymarket buys between 5 and 20 cents of at least $10,000 with holdout. If it holds, sweep the top price from 10 to 30 cents and show where the return peaks."
Explain what just moved
"Show me the biggest market moves of the last hour and, for the top three, who was buying and at what price."
Look into one market
"Look into the Fed's next rate decision: the prices, who holds each side, and how those traders have done on Fed markets."
Run an idea on paper
"Start a paper bot that copies $50 a trade of whale buys on tennis favorites under 75 cents, and tell me how it is doing at the end of the week."

Every tool and the scope it needs

A key only ever sees the tools its scopes allow. A read key gets every read tool; write adds paper orders and paper bots. No tool moves money.
Tool	Group	What it does	Scope
get_trades	Whale money	Large trades, newest first, reaching back 30 days, with the market, side, price, dollars and the named trader's record. Each row older than an hour counts toward the 5,000 trades a day.	read
get_positions	Whale money	What traders with a winning record hold now, biggest first. Each holder counts as one row.	read
get_moves	Whale money	Sharp moves, favorite flips and whales piling in over the last day. Pass since to read only what is new.	read
search_markets	Markets	Markets by name or activity, open, settled or both, with refs for the other tools.	read
get_market	Markets	One market, with optional book, whales, call, history and trades. Each include costs one more request.	read
search_traders	Traders	Traders ranked by profit, win rate, return, volume or activity: the top 100 for any ranking. Each call counts toward the 10 trader listings a day.	read
get_trader	Traders	One trader's record, with optional positions, curve, habits and similar traders. Each different trader counts toward the 100 traders a day.	read
backtest	Testing	Score copying a signal or a trader on settled markets, with optional holdout, sweep and breakdown. Places nothing.	read
place_paper_order	Paper	Buy or sell one side on paper at the live price. Pass a requestId so a retry never places it twice.	write
get_paper_account	Paper	Paper positions placed by hand and by bots, with profit.	read
list_bots	Bots	Every bot on the account with its record.	read
get_bot	Bots	One bot's settings and record, with optional trades and activity.	read
create_bot	Bots	Start a paper bot that copies a trader, a signal, or trades one market by a price rule.	write
update_bot	Bots	Change a paper bot's settings, or pause and resume a trader or market bot.	write
delete_bot	Bots	Stop and delete a paper bot. Its record is kept.	write

Built for an assistant

Paper only. Paper orders and paper bots trade at live prices and no money moves.
Only the tools a key can use. A read key never sees the paper tools, so a model never plans around a call it cannot make.
Safe retries. A paper order carries a requestId, so a retry after a timeout returns the first order instead of placing a second.
Errors that say what to do next. Every failure carries a stable code and a hint.
Answers sized for a model. Lists return 20 rows unless asked for more, and every tool says which one to call next.

Limits

Frequently asked questions

What is a prediction market MCP server?

Is it local or remote?

Remote, over HTTP. There is nothing to install or run. Add the endpoint to your client with your API key as a bearer header and the tools appear.

Which clients work?

Can an AI spend my money through this?

What does it cost?

How is a backtest's return worked out?

Is there a plain REST API too?

Yes. Every tool returns the same data as a v2 API route, with the same keys and limits. Use the API when you want to script it yourself and MCP when you want an assistant to drive.

Give your assistant the tape

Whale trades, trader records, backtests and paper bots in Claude, ChatGPT or Cursor. Included with Rivo, $9 a week or $15 a month. Just one trade can pay for it.

Get access

Prediction Market MCP Server

Give Claude, ChatGPT or Cursor the whole prediction market tape

Ask your AI what winning traders are buying, then test it

Winning traders your assistant can open right now

Connect: add Rivo to your assistant in a minute

What you get

Read the tape: see what winning traders are doing right now

What you get

Judge traders and ideas: backtest before you bet

What you get

Run it on paper: turn an idea into a bot without spending a dollar

What you set

Prompts that know the order

Things to ask it

See what winning traders hold

Shortlist Kalshi traders worth copying

Test a longshot idea

Explain what just moved

Look into one market

Run an idea on paper

Every tool and the scope it needs

Built for an assistant

Limits

Frequently asked questions

What is a prediction market MCP server?

Is it local or remote?

Which clients work?

Can an AI spend my money through this?

What does it cost?

How is a backtest's return worked out?

Is there a plain REST API too?

Give your assistant the tape

Related

Prediction Market MCP Server

Give Claude, ChatGPT or Cursor the whole prediction market tape

Ask your AI what winning traders are buying, then test it

Winning traders your assistant can open right now

Connect: add Rivo to your assistant in a minute

What you get

Read the tape: see what winning traders are doing right now

What you get

Judge traders and ideas: backtest before you bet

What you get

Run it on paper: turn an idea into a bot without spending a dollar

What you set

Prompts that know the order

Things to ask it

See what winning traders hold

Shortlist Kalshi traders worth copying

Test a longshot idea

Explain what just moved

Look into one market

Run an idea on paper

Every tool and the scope it needs

Built for an assistant

Limits

Frequently asked questions

What is a prediction market MCP server?

Is it local or remote?

Which clients work?

Can an AI spend my money through this?

What does it cost?

How is a backtest's return worked out?

Is there a plain REST API too?

Give your assistant the tape

Related